In recent years, double-deck elevator systems (DDES) where two cars are connected in a shaft has been developed for the rising demand of more efficient transport of passengers in high-rise buildings. DDES has specific behaviors due to the connection of cars and the need for securing comfortable riding, so its group control becomes more complex than conventional single-deck elevator systems (SDES). Meanwhile, a graph-based evolutionary method, Genetic Network Programming (GNP) has been applied to elevator group supervisory control systems, and its effectiveness is clarified. GNP can consider the specific behaviors of DDES in its node functions easily and execute an efficient rule-based group control optimized evolutionary. In this paper, a new group control system for DDES using GNP is proposed, and its optimization and performance verification are done through the simulations. First, optimization of GNP for DDES is executed. Second, the performance of the proposed method is verified by the comparison with conventional methods, and the obtained control rules are studied. Finally, the performance improvement by the proposal is evaluated in terms of SDES capacity.